In the latest release, the “append” function of Array is deprecated, then what is the good way to “append” a new item?
the notice is :
/main.mo:3017.20-3017.32: warning [M0154], field append is deprecated:
Array.append has critical performance flaws; use a Buffer, and Buffer.append, instead.
but there is no filter/find functions for Buffer, which means I have to convert buffer to Array for filtering? or any other better ways?
Believe it or not, but many of the array class methods convert the array to a Buffer, and then back to an Array.
See the Array.filter implementation in motoko-base - I would recommend just copying the Buffer filter logic/functionality from there. You can expand upon that to do the same for .find().
Buffer has .get() methods for retrieving indexes as well, so I would honestly only use an Array if you’re sure that your index-able list isn’t going to change in size - otherwise, Buffer all the way.
Hello, so just for clarification, converting the array to a buffer is straight forward enough, but unless I am missing something a Buffer is not a stable variable type … what would be the recommend path if you want to have a stable var outcome?
Would you convert the array to a buffer and back again to maintain stability? or build it as a buffer with upgrade hooks to put them to arrays and back to buffers after? like one might for a hashmap? … just looking for a general consensus for lowest denominator on performance and best practices. … if you have an app based on stable arrays what would you do?
ie. if I have the following:
return Array.append<Project>(theProjectsNew, [newProject]);
which appends a new object “newProject” to an array called “theProjectsNew”
and replace it with:
let theProjectsBuffer : Buffer.Buffer = Buffer.Buffer(theProjectsNew.size());
for (x in theProjectsNew.vals()) {
theProjectsBuffer.add(x);
};
theProjectsBuffer.add(newProject);
return theProjectsBuffer.toArray();
the programatic results are the same … and the warning goes away … is this a valid way forward? and leave things otherwise as stable arrays?
You are correct that Buffer is not a stable variable type, because the Buffer class is an object that containing local functions. According to the documentation,
“Like shared types, stable types are restricted to first-order data, excluding local functions and structures built from local functions (such as objects).”
Let’s take a second to look at the Buffer class signature:
public class Buffer<X>(initCapacity : Nat) {
var count : Nat = 0;
var elems : [var X] = [var]; // initially empty; allocated upon first `add`
...local functions
}
Although the Buffer class is not stable, the count and elems local variables are stable, as long as the X type involved passed is stable. This means you can take the add method directly from Buffer.mo, rip it out of the class, and turn it’s signature from:
public func add(elem : X)
to
public func add<X>(elems: [X], count: Nat, elem : X)
I had been meaning to write a stable buffer module, so you’re comment pushed me to do it here 
- https://github.com/canscale/StableBuffer.
Please feel free to use it, as well as any of the other stable-adapted modules I’ve written (StableHashMap, StableRBTree)
You can do this conversion back and forth if you want to keep things stable, but it’s pretty inefficient IMO. Better just to keep it as a buffer if you don’t have a fixed bound on the size of the array.
You can keep doing this conversion (stable, but inefficient), or use a stable buffer like this library I mentioned in my other post https://github.com/canscale/StableBuffer. If you don’t need direct lookups by index, Lists are a great solution.
In the code above, you are adding all the elements in the array to the buffer.
And in this code (directly above), you are essentially creating an array and adding all the elements from the buffer back to the array.
See the toArray() implementation I’ve copied in below.
// https://github.com/dfinity/motoko-base/blob/master/src/Buffer.mo#L105
public func toArray() : [X] =
// immutable clone of array
Prim.Array_tabulate<X>(
count,
func(x : Nat) : X { elems[x] }
);
Basically, always use buffer and convert it to an array preupgrade and back to a buffer in your init function(or post upgrade.).
@icme - Totally makes sense to me … I will also take a look at that StableBuffer and StableHash map repos … Thanks again for the detailed reply … will work a refactored solution in our platform over the next week… the amount of data in play at the moment is negligible so surviving the “deprecation” of Array.append was our priority
@skilesare - thanks as well … I was actually unaware of the pre/post upgrade options when I started our project a bunch of months ago, but will factor it as an option for this application for sure.
Updating the solution to use this permalink
since the original solution has so many clicks and now points to the wrong line of code.